R/lagged_correlations.R

In pawel-wieczynski/PolishStock: Analysis of WSE stocks and other assets.

#' Lagged Correlations function
#'
#' Calculate correlation beetwen two numeric vector x and y for lagged values of x. Both vectors need to have equal length.
#' @param x Numeric vector of asset prices or returns (to be lagged)
#' @param y Numeric vector of asset prices or returns
#' @param max_lag Max number of lags for vector x. Default set to lengt(x) - 2
#' @param plot If TRUE plot correlation for each lag
#'
#' @return Numeric vector of correlations for different time lags.
#' @export
#'
#' @examples
#' tickers <- c('wig20', 'mwig40')
#' sapply(tickers
#'        ,stooq_download
#'        ,destination = 'datasets\\')
#' files <- list.files('datasets\\')
#' close_prices <- df_prices(files = files
#'                           ,source = 'datasets\\')
#' x <- close_prices[,2] %>%
#'   unlist()
#' y <- close_prices[,3] %>%
#'   unlist()
#' lagged_correlations(x, y, max_lag = 200)

lagged_correlations <- function(x
                               ,y
                               ,max_lag = length(x) - 2
                               ,plot = TRUE){

  if (class(x) != 'numeric'){
    stop('x is not numeric vector')
  }

  if (class(y) != 'numeric'){
    stop('y is not numeric vector')
  }

  if (length(x) != length(y)){
    stop('lengths of x and y differ')
  }

  rho <- c()
  rho[1] <- cor(x, y)

  for (i in 1:max_lag) {

    rho[i+1] <- cor(
      as.numeric(
        na.omit(dplyr::lag(x, i))
      )
      ,y[-(1:i)]
    )

  }

  if (plot == TRUE){

    plot(1:length(rho)
         ,rho
         ,type = 'l'
         ,lwd = 1.5
         ,xlab = 'Lag'
         ,ylab = 'Correlation')

  }

  return(rho)

}